EMI Measurements: Methodology and Techniques...EMI transient signals can have peaks at either...
Transcript of EMI Measurements: Methodology and Techniques...EMI transient signals can have peaks at either...
EMI Measurements: Methodology and Techniques
Vladimir Kraz
OnFILTER, Inc. 3601-B Soquel, CA 95073 USA Tel. +1-831-824-4052 email [email protected]
Abstract - This paper describes some aspects of methodology, instrumentation and techniques of
measuring high-frequency electrical noise (EMI) in electronic manufacturing environment. High-
frequency measurements are quite different from typical ESD-related measurements which this paper
explores.
Introduction High-frequency noise (often, albeit technically
incorrectly, called "EMI") is one of the sources of electrical overstress (EOS) that damages sensitive
components. It also causes malfunction of
electronic equipment.
A parameter that cannot be measured cannot be
controlled. Measurements of high-frequency
signals differ radically from measurements of
traditional ESD parameters. This paper will
present methodology and techniques for some
common high-frequency measurements in
manufacturing environment as well as outline
critical properties of the EMI and EOS signals as
related to a manufacturing environment. Many
details are omitted due to limitations of space.
The basic differences between
EMI and ESD measurements Most ESD measurements deal with DC or
extremely low-frequency signals of very high
impedance. Examples include static field or voltage, high resistance values of dissipative
materials, ionization balance and alike. This is
exactly opposite to measurements of high-frequency signals which require high bandwidth,
often to up to gigahertz range, and, especially for
conducted measurements, matched low impedance
of the instrument. Use of common-type ESD
instrumentation or other general-purpose tools
produces not only inaccurate results, it simply
cannot produce any relevant results at all. As an example, consider ubiquitous static field meter.
Its vibrating sensor works at a frequency of
~300...600Hz and after signal processing the
theoretical bandwidth of 1/2 of that frequency
ends up to be of less than 10...15Hz. The refresh
rate of the display further limits it to ~3Hz or less. Feedback-operated instruments, such as certain
static voltmeters and CPMs have similarly low
bandwidth. As a reference, conducted EMI on power lines starts at several tens of kilohertz and
radiated emission of consequence starts at low
megahertz. Common multimeter or current clamp
have typical bandwidth of up to several kilohertz
with the best of them up to few hundred kilohertz,
which is grossly insufficient. Thus, a completely
different set of tools and a very different methodology needs to be used by an ESD
specialist for analysis of EMI environment.
Fundamentals of High-
Frequency Signal Metrology
Measurement Types
High-frequency signal measurements can be of
time domain, frequency domain, or broadband.
Frequency domain
Frequency domain measurements while incapable
of discerning the waveform of the signal, are
capable of providing the spectral characteristics of the signal. Most of applications for frequency
domain measurements are in EMC compliance
tests, wired and wireless communication and
broadcast. Frequency domain measurements are
performed with specialized instruments - spectrum
analyzers - and to a lesser degree using FFT
function of some oscilloscopes.
Time Domain
In time domain the waveform of the signal is
studied and a variety of parameters are collected,
among them are peak, average or RMS values, rise
and fall times, energy, repetition rate and others.
Time domain measurements are especially suited
for transient signals, such as artifacts of ESD
Figure 1. The same signal measured with matched (a) and
mismatched (b) impedance
Figure 2. Signal reflections due to boundary conditions
Events, noise on power (AC and DC) lines, ground
and alike. Time domain measurements are made
with an oscilloscope. Most high-frequency
measurements in electronic manufacturing
environment are time domain.
Broadband
Broadband measurements provide data only on one parameter of the signal - its amplitude
regardless of signal frequency, waveform and
anything else. A ubiquitous multimeter is a
broadband instrument.
Since time-domain measurements are prevalent for
electronic manufacturing environment the rest of
this paper focuses of this type of measurements.
Coaxial Cables
Essential test accessory for measurement of high-
frequency signals is coaxial cable of the right type.
Coaxial cable provides two benefits for high-
frequency measurements: shielding of often weak
measured signal from outside influence and
elimination of induced loop currents that would be
present in case of conventional test leads spread
apart. Of course, a proper coaxial cable need to be
selected. Among the parameters to watch for are
cable impedance and attenuation at high
frequencies within the band of interest. A note of
caution: coaxial cables are fragile by nature.
Bending and kinking them will irreversibly
damage them causing signal distortion.
Impedance Matching
At high frequencies the impedance matching is a
must for most cases - impedance mismatch leads
to signal distortions and ringing. The most typical
output and input impedance of high-frequency
instruments is 50 Ohms, although in some cases,
such as TV cable, it may be 75 Ohms. Figure 1
shows the same signal measured with matched
impedance (50 Ohms) and mismatched impedance
(an oscilloscope set to 1M). As seen, mismatched
impedance causes significant ringing of the signal
and also readings of higher peak amplitude.
A good example of a mismatch is measurements of ESD Events using a passive near field probe
(usually "ball" or "stick" antenna) and an
oscilloscope as described in [1] and [2]. Such
passive near-field probes which are, in essence,
small antennae, may have 50 Ohms impedance
only at some particular frequency. Since an ESD
Event signal has wide spectrum, there will be plenty of frequencies at which the output
impedance of an antenna is not 50 Ohms and this
causes significant ringing.
Ringing due to Boundary Effects
Reflections from the edges of board or other metal
surface also cause ringing - ultimately, it is also an
impedance mismatch. Figure 2 shows how the
reflections at the edge of the PC board occur. This
document provides good examples of ringing
caused at least in part by the boundary reflections
and how the timing of the reflections is defined by
the size of the board [3].
Not all ringing of ESD Event signals is artifacts of
measurements - some are genuine signal, but
separating it from the measurement errors need to
be meticulously done.
Figure 3. Log-periodic antenna and its
antenna factor
Figure 4. Comparison between regular passive antenna
and a special flat-frequency active antenna
Frequency Response
Radiated emission is measured using antennae.
Most antennae do not have flat frequency
response. Many are tuned to a specific frequency
or band. Even broadband antennae, such as log-periodic
ones used
for EMC
test may
have very
uneven
frequency response
offering up
to more than 20dB
(10 times)
difference
in
sensitivity
(Antenna
Factor - AF) within
the band of interest (see Figure 3) [4]. When the
measurements are done using spectrum analyzer
which measures one frequency at a time, a
correction factor can be applied for each
frequency. However, no correction can be applied
for time domain measurements since all frequencies are measured at once. The result is
distorted waveform that misrepresent the amount
of energy of the discharge. Figure 4 shows
comparison between frequency responses and
transient responses of a regular passive antenna
(equivalent to the abovementioned "ball" antenna)
and a specially-designed active antenna for ESD
measurements with the flat frequency response. As seen, the regular frequently-used passive antenna
has very poor response at low end of the spectrum.
This results in apparent absence of energy
component of an ESD Event and misjudgment of
its effect. A note: this active antenna has an
inverting amplifier inside, therefore the polarities
of the picked up signal appear to be opposite.
Oscilloscope Triggering
A typical oscilloscope triggers on either positive
or negative threshold, but not on both. However,
EMI transient signals can have peaks at either
polarity. It is important to do signal capture using
trigger on both polarities - positive and then
negative. Neglecting any one polarity may cost
you missed signal.
Many EMI events are multiple. If Single mode of
triggering is used, only first such event is captured.
If Normal mode is used, it would the display the
last captured event. What happens between these
events is thus lost. Some digital storage
oscilloscopes have memory that can store more
than one captured event which helps. Where the problem rises is in trying to observe both the
waveforms of each event and the interval between
them on one screen to understand the timing. The
duration of an ESD Event may be only few
nanoseconds but the spacing between multiple
discharges may be several milliseconds or longer.
Screen resolution is not sufficient to display such short events if several of them are captured in one
shot. In such cases specialized monitors and data
acquisition system may be required.
Multiple non-ESD EMI Events are much easier to
capture and observe because such events have
longer duration and their repetition rate is in the
same range as the event's duration.
Differential Signals and Ground Loops
Measurements of conducted emission with the
scope are often made using regular oscilloscope probe. The problem is that ground contact of the
probe is connected to mains' ground via power
cable of the scope. This creates ground loop
because ground in the electrical outlet is not
necessarily, and most likely not, ground vs. which
the measurements should be made. In addition,
this may add noise coming from the mains' ground
to the measured signal.
There are several ways to avoid the above
problems. One is to use battery-powered
oscilloscope. This would resolve ground loop
Figure 6. Fundamental properties of ESD Event
Figure 7. Typical noise on power
lines caused by light dimmer
Figure 5. Special power line EMI Adapters
problem, but will leave in place a problem of
capacitive coupling between the chassis of the
scope and grounded surfaces nearby. Another way
is to use either a differential probe or two probes
on two channels and use "A-B" function. This
would eliminate the parasitic capacitance problem.
A third way is using special power line EMI
adapters (Fig. 5) which provide balanced input and
have single-channel output. They also protect
oscilloscope from the potentially damaging
voltages on measured conductors, such as power
lines.
Measurements of noise on power lines
Noise on power lines of all kinds is the prime source of EMI in manufacturing environment. The
biggest challenge of this type of measurements is
that the signal of interest, i.e. high-frequency noise is significantly smaller than the AC mains voltage.
In addition to not being able to see the desired
signal, this also causes triggering problems.
Another challenge is that high voltage on mains
(250V RMS corresponds to 353V peak and
required scope range of at least 700V on the
screen. The solution to this problem is special power line EMI adapters which perform several
functions: complete blockage of AC mains voltage
while passing through high-frequency signals without alteration; balanced input and overload
protection.
Waveform Properties Once the waveform is captured by a high-speed
digital storage oscilloscope, it needs to be properly
analyzed. Depending on the purpose of the
measurements sometimes different parameters
may need to be considered.
ESD Events
The three critical parameters for analysis of
properties of ESD Events as shown in Figure 6 are peak amplitude (either polarity), rise time and
energy (area under the curve) [5]. Peak amplitude
is a function of maximum discharge current
through the device. Rise time indicates how fast
the discharge energy flows into the device - the
faster the worse for the device. "Area under the
curve" of the pulse is an indication of how much
energy was injected into the device. Another
parameter to consider is whether multiple
discharges are taking place (a frequent occurrence). This is important for two different
reasons. One is that closely-spaced multiple
discharges add to the energy of the discharge and
may have cumulative effect; another is that
measurements of multiple discharges requires
special equipment and special techniques.
EMI of non-ESD Origin
High-frequency noise in manufacturing environment is largely transient signals and
repeatable
pulsed signals. The waveform
of the signals is
quite different
from the ESD Events. There
is no sharp rise
time; the ringing is frequently
present and the
Fig. 8 Ground Current in Servo Motor
Figure 9a. Ground Current in Servo Motor without
Filter: 1.12A
Figure 9b. Ground Current in Servo Motor with the
Filter: 0.0186A
Figure 9c. Measurement Setup
signal is often periodic (Figure 7). The repetition
of EMI signal is largely synchronized with one of
the following:
• AC mains. Note that the repetition frequency is
twice the mains frequency, i.e. 100Hz for a
50Hz system and 120Hz for a 60Hz system.
• Switching power supply frequency. This
typically ranges anywhere from 40kHz to
1MHz
• Servo motors. This frequency is often ranges
from 8kHz to 20kHz.
There are also plenty of non-synchronized
occasional EMI signals caused by turning on and
off lights, motors and other electrical equipment,
as well as operation or end switches and solenoids.
For the narrow scope of electronic manufacturing
environment, the following parameters are of
interest: peak amplitude (both positive and
negative), rise and fall times and energy (the area
under the curve of the waveform). All these
parameters except the last one are easily
obtainable from the scope itself. The energy needs
may need to be calculated from the collected
captured waveform data. This is where flat
frequency response of an antenna makes a
difference between correct and incorrect data.
Real-Life Example:
Servo Motor Noise Semiconductor manufacturing tools such as wire
bonders, IC handlers and alike use servo and
variable-frequency motors. These motors utilize
pulses that generate significant currents in tool's ground which, in turn, can cause electrical
overstress (EOS) in the devices. Two types of
measurements can be made - voltage and current.
Figure 8 shows current measurements in ground
synchronized with the rise of the drive pulse.
Current measurements were made with Tektronix'
CT1 current probe.
Figure 9 shows reduction of ground current after connecting OnFILTER' servo motor filter between
the servo amplifier and the motor.
As seen, the EMI and a possible EOS problem was
easily identified, the mitigating measures were
implemented (special servo motor filter) and the
improvements were verified.
Conclusion Proper measurement methodology and techniques
are essential for effective management of EMI
environment in manufacturing environment. Understanding of properties of high frequency
signals and specifics of their measurements help
ESD specialists at the factory to be successful in
this endeavor.
References [1]Unifying Factory ESD Measurements and
Component ESD Stress Testing, J. Montoya et.
al., ESD Symposium Proceeds, 2005
[2] Electrostatic Discharge Detection: Antenna
and Oscilloscope, Intel instructional video,
http://www.intel.com/content/www/us/en/quality/e
sd-detection-antenna-and-oscilloscope-
video.html?wapkw=ball+antenna
[3] Investigation on Discharge Current Waveforms
in Board-Level CDM ESD Events With Different
Board Sizes" Yuan-Wen Hsiao and Ming-Dou
Ker, International ESD Workshop 2008
[4] Inside EMC Antennas, Tom Lecklider, Feb.
2010
[5] Verification of ESD Environment in
Production, V. Kraz, Il Controllo dell'Elettricita
Statica, Milano 29 Ottobre 2002